The history of soil mapping in Denmark shows how the evolution of soil classification systems and the gradual incorporation of additional soil and environmental data helped to improve the knowledge on Danish soils. The work on the first soil inventory in Denmark started in 1688 when the King called for the Land Register to provide the basis for taxation. The surveyors, therefore, evaluated the land based on potential crop yields rather than specific soil characteristics. Four experienced farmers were responsible for evaluating the land in each parish. The surveys did not use standardized methods, so the classification varied between administrative regions. In Jutland, which comprises the largest part of Denmark, the soils were classified ranging from “most fertile” to “poorest”. According to this classification, a certain number of area units (1 tønde land [tdl] = 0.53 ha) was associated with the tax value of one barrel of “hard grain” [th] (i.e., barley or rye). For instance, the poorest soils had a value of 1 th per 16 tdl, while the most fertile soils had a value of 1 th per 2 tdl. The register did not map the classes but instead listed the values for the fields in each parish.

In the 18th century, the Land Register produced the first soil maps at 1:4,000 scale for two regions in Denmark, and the land evaluation now incorporated soil characteristics, such as “good mull and sand for rye and barley”. These new surveys laid the basis for the Great Danish Land Register of 1844, which resulted in a national land value map for taxation (20). The Great Danish Land Register was a milestone in soil inventories, in which the land evaluation explicitly considered soil characteristics in addition to yield potentials. The new surveys adopted a characterization system of grades from 0 to 24. The “optimal soil” had a grade of 24 and represented a reference of suitability for farming in wet and dry conditions. This reference soil, located in Karlslunde just southeast of Copenhagen, had a thick topsoil (> 31 cm) with a mixture of clay and mull, a clayey subsoil, and a gentle south-facing slope. Later, four soils with grades 20, 16, 10, and 3 were added to the reference list of “normal soils”. The soil evaluation relied on a combination of factors described by the surveyors in the field. The land register formed the basis for taxation until the end of the 19th century. However, due to agricultural advances and management (e.g., tile drainage), the productivity of the land started to diverge from the assigned values. As a consequence, from 1903 onwards, new tax legislation used the commercial value of the land as a replacement (18).

A new National Land Assessment was conducted in 1949 to provide a comprehensive overview of the agricultural lands in Denmark. The soils were rated on a 0–100 point scale for their natural characteristics and crop yields, and mineral soils were classified into seven classes ranging from heavy clay (50–90 points) to gravelly sand (0–30 points) (21). The rating was partly based on the presence of mull, water retention, drainage conditions, soil colour, and the presence of plant roots. Despite advances in describing soil characteristics, soil sampling and classification still relied on human judgment and support from experienced farmers. The surveys comprised assessments from more than 25,000 locations and enabled the production of top- and subsoil textural maps. In 1970, the Land Assessment Commission recommended a new national survey, but the government considered the plan too expensive.

However, in 1975, the Danish government realized that the information on soil in agricultural land was insufficient for effective land use planning. This motivated the introduction of the Danish Soil Classification (DSC), and between 1975 and 1980, a new campaign collected about 36,000 soil samples (the DSC database) and produced maps of textural classes at a scale of 1:50,000 (18, 22). For the first time, the survey used laboratory measurements of texture, soil organic carbon (SOC) and CaCO₃. The data was stored in databases and a GIS-based soil information system was developed (23, 24). The classification system comprised 12 soil types, later simplified to eight colour codes based on soil texture and organic matter and calcium carbonate content. The soil classes were manually delineated based on knowledge of soil variation from local agricultural advisors and supported by detailed geological surface maps. The soil survey also published 320 paper map sheets at 1:50,000 scale. These maps also provided information on the surface geology, slope and drainage.

Additional inventories assembled in the 1980s further contributed to the knowledge on soil variability in Denmark (25, 26). During the construction of a national gas pipeline between 1981 and 1984, about 8,500 soil profile investigations were carried out along two transects [NSGP database (19)]. In the same period, a survey of potentially acid sulfate soils collected samples from more than 8,000 augering sites [Ochre Classification database (27)]. Lastly, in 1987−1989, a nationwide programme to study nitrogen (N) dynamics described and sampled about 850 soil profiles located in a 7-km grid [Danish Soil Monitoring Grid or DSMG database (18, 22)]. As the DSMG forms the basis for a monitoring programme for SOC, subsequent campaigns have collected augering samples in 1997, 2009 and 2019 (28).

Until the 1990s, most soil investigations focused on agricultural areas. However, in the beginning of the 1990s, a Danish Forest Site Classification system was developed for mapping soil prior to afforestation to ensure the long-term health of the forest. The surveys included one auger boring per hectare and a soil profile description for each soil type present in the area (29). Later soil surveys for afforestation projects have also employed the classification system developed for forest soils [e.g (30)]. As a consequence of the Kyoto protocol, a national soil survey of Danish peatlands (SINKS) was initiated, and about 10,000 auger samples were originally collected between 2009 and 2010 in order to map the distribution of organic wetland soils (31). Under the European Union (EU) structure, soil samples were also collected across Denmark to improve the LUCAS database over time (32). Also, from Denmark’s many soil profile- related experiences, the nation took a leading role in the design of a European Soil Profile Analytical Database (SPADE 18) and a harmonized soil map for the EU at a scale of 1:1,000,000 (33). As a result of several soil survey programmes, Denmark has one of the most representative (~ 2 soil sampling locations per km²) and detailed national soil databases in the world. The Danish soil information is presently under a process of harmonization following the existing regulations and directives for Open Databases (https://inspire.ec.europa.eu/data-specifications/2892).

The use of proximal sensors opened up for the opportunity to obtain more information about soil characteristics [e.g (34)], and visible near-infrared (Vis-NIR) spectroscopy has been successfully applied, showing great potential in supplementing wet chemistry for estimating soil properties in Denmark (35). The Danish Vis-NIR spectral database consists of measurements from about 7,000 soil samples from the DSMG investigations (1986 and 2009), and from surface and profile samples from several field surveys, thus covering the different soil types representative for the country and different spatial scales. Data was gathered using different sensor types in the laboratory and in the field, including bench-top, portable and on-the-go systems.

Denmark started to monitor the environment to understand and manage problems related to agricultural areas and water pollution, and later to protect natural habitats. This resulted in spatial and temporal representative environmental datasets related to land use and management, water dynamics quality, and land surface characteristics at high resolution.

Agricultural lands cover about two thirds of Denmark, and information about land use and crops has been available at the regional level since 1848. Denmark joined EF (European Community) in 1972 and the Common Agricultural Policy in 1996. Field blocks were introduced in Denmark − a dataset containing information on the maximum eligible agricultural land use as defined by the EU. Field blocks are digitized GIS polygons, and a block encompasses one to approximately 15 fields delimited by stable landscape structures such as roads and fences (36, 37). This dataset was further detailed in 2011, when farmers started to register the field boundaries and crop types in each field every year (within a block). This resulted in a detailed land use database, which has the potential to be used in several research areas. For instance, this dataset was one of the main components in the creation of the present land cover map for Denmark (38). Furthermore, national surveys with the Light Detection and Ranging (LiDAR) sensors in 2006/07 (1.6-m), 2014/15 (0.4-m), 2018/22 (0.4-m) provide information on topography and vegetation at high resolutions (39). The LiDAR-based DTM products (DHM Terræn) are freely available (https://dataforsyningen.dk/). Recently, drones with high-resolution spectral cameras and proximal sensors have started to be employed to evaluate land, vegetation dynamics, and sub-surface soil properties at high resolution.

The Danish government funded several programmes to monitor water quality and dynamics to inform protection efforts for the extensive coastline as well as groundwater, which is the main source of drinking water. A national groundwater monitoring programme (GRUMO) was established in 1988, covering 74 well catchment areas and six small agricultural catchments with more than 1,500 screens (i.e., portions of the well that serve as water intake at different depths) (40). The water is analysed in these locations, mostly annually, for main components, inorganic trace elements, organic micro-pollutants, and pesticides. Additional water samples from about 10,000 wells across Denmark are analysed every three to five years. Relevant parameters on geology, groundwater quality and water levels at wells are collected in the national open-access well database, JUPITER [https://data.geus.dk/JupiterWWW (41)].

The lack of knowledge about forests motivated the creation of the Forest Tree Species Trial Plots in the 1960s, with each containing even-aged single-species trees distributed in clusters across 13 sites with different characteristics. The 13 sites were used to compare growth and health of 12 tree species in typical site types across Denmark: oceanic climate and sandy outwash plains in the west, continental climate and clayey soils in the east, and intermediate climate and gravelly tills in the central regions (42). The studies on tree species increased and a total of 2,130 forested plots were measured during the Danish National Forest Inventory (NFI) in 2006/2007 (43). The Danish NFI is based on a 2 × 2 km grid across the Danish land surface. In each grid cell, a cluster of four circular plots for measuring forest factors (e.g., growing stock, biomass, and total forest C stock) is placed in the corners of a 200 × 200 m square.

Understanding the water dynamics in the soil is crucial for the economy and human well-being, especially in Denmark, where artificial drainage is a prerequisite for agricultural production in many places. At the same time, the groundwater provides clean water for consumption and is a source of irrigation in areas with sandy soils. Artificial drainage played a crucial role during the agricultural expansion and intensification in the 19^th century in Denmark and the cultivation of wetland soils in the 20^th century (60). Today, about 50% of the arable land in Denmark is artificially drained (61).

Agricultural drainage, combined with the increasing use of fertilizers and pesticides in the 20^th century, increased the risk of groundwater contamination and eutrophication in surface waters. In the 1980s, these risks created a need to monitor water quantity and quality (62). New monitoring programmes therefore aimed to collect information on water dynamics in streams and groundwater to assess the risks of nutrient and pesticide leaching. The efforts included a comprehensive network of wells and gauging stations to collect information on water quality and flow. However, these monitoring efforts did not provide sufficient information for an understanding of the nutrient transport dynamics in the soil. This need motivated the establishment of the national DSMG programme (Section 2), and additional programmes focused on water permeability in representative soils across Denmark (63), although his data was available only for specific sites. With the development of DSM, it became possible to map water dynamics in more detail. New studies therefore aimed to map natural soil drainage (64, 65), artificially drained areas (66), macropore flow (67), interactions between the groundwater and surface waters (68), and variations in the groundwater table (69).

Møller et al. (64) mapped the five soil drainage classes for Denmark using information from 1,702 soil profiles from the gas pipeline investigation and DSMG database in a decision tree model. Later, Beucher et al. (65) used artificial neural networks to map soil drainage classes and found that their predictive accuracy was comparable to decision tree classification. Moderately drained soils are the dominant drainage class in the loamy till areas in eastern Denmark, while well-drained and very well-drained soils mainly occur in the sandy upland soils in the west.

Although soil drainage classes give a good indication of the expected water saturation in the soil, artificial drainage often overshadows the patterns of natural soil drainage. Olesen (61) compiled 745 field observations of the presence or absence of artificial drainage across Denmark. Møller (66) used this dataset and an ensemble of 77 machine learning models to map artificially drained areas for Denmark and the associated uncertainties. The models predicted a high presence of artificial drainage in the loamy tills of eastern Denmark and in the large wetland areas of northern Denmark. However, even with this information, the location of the individual drainage pipes is still often poorly documented or unknown. The information on drainage pipe locations can help farmers repair damaged pipes and install new systems, but it can also inform the establishment of mitigation measures to remove nutrients, pesticides and pathogens from the drainage water. Conventional methods for mapping drainage pipes include probing or the use of trenching equipment, which are time-consuming and invasive. Several studies assessing the use of proximal soil sensors were conducted in Fensholt, Denmark ( Figure 4D ). In particular, Koganti et al. (70) showed that ground-penetrating radar (GPR) was potentially useful to localize drainage pipes, while a magnetic gradiometer proved ineffective. Furthermore, proximal sensors have a limited capacity to cover large areas, and efforts are therefore underway to test the suitability of unmanned aerial vehicle (UAV) imagery for this purpose. Koganti et al. (71) found that the best approach was to use both GPR and UAV techniques when possible, as they can provide complementary information for mapping drainage pipes. They also provided recommendations for the timing of sensor surveys, while they argued that future research should focus on developing a formal framework with specific guidelines for different soil types.

The drainage pipes increase the volume of water transported to surface water bodies, and predicting the drainage discharge could be an important source of information to analyse the risks of water contamination (72). Using machine learning, hydrological data from 53 drainage stations across Denmark and several explanatory variables, Motarjemi et al. (73) predicted the national-scale tile drainage discharge. They found that precipitation, topographic elevation, and clay contents below tile drains were the most important explanatory variables. However, this study did not take into consideration the variation of the predictor variables at the catchment scale.

An important function for the soil is its capacity to filter nutrients and pesticides from the water that moves through it. This filtering process depends very strongly on the flow paths of the water. For instance, preferential flow through macropores can cause P and pesticides to appear in the groundwater (74). Using pedotransfer functions (PTFs), soil texture maps and data from several locations in Denmark, Iversen et al. (67) mapped the risk of macropore flow in Denmark. The availability of soil texture maps in raster format (50) therefore directly enabled the spatial prediction of macropore flow. The higher resolution of the soil textural maps from 2013–2014 and robust PTFs enabled the development of maps of saturated and unsaturated hydraulic conductivities as well as water contents at specific pressure heads [ Figure 4B (75)]. These results indicated that soils in eastern Denmark, which contain higher amounts of clay, have a higher probability for macropore flow.

Soil texture also affects groundwater-surface water interaction (GSI), which occurs when groundwater leaks into surface water or surface water percolates into the groundwater. GSI helps manage water resources and protect land and aquatic environments, essential for sound policymaking (76). Danish researchers reviewed GSI studies to create an eco-hydrological typology (77). Landscape, riparian hydrogeology, and flow path types were classified on arbitrary catchment scales of >5 km, 1–5 km, and 10–1000 m stream lengths. This set the stage for spatial classification mapping by first delineating Denmark’s river valley bottom (78). Sechu et al. (68) then developed a mapping routine to classify the Danish stream network using the typology. Three GSI contact types were mapped ( Figure 4A ): connected clayey, connected sandy, disconnected; and riparian flow paths: diffuse/overland, direct, and artificial drainage. The resulting maps revealed that 85% of Danish streams are connected to groundwater in sandy and clayey subsoils, and 87% receive riparian water through direct flow paths. For the stream sections, 41% and 19% receive riparian water through artificial drainage or diffuse/overland flow. These numbers match Denmark’s land use with 61% under agriculture (79), limiting runoff, and 50% artificially drained (61). The maps improve our understanding of flow of water and nutrients from uplands to streams and can help decision-makers manage agricultural land to protect receiving waters.

The combined water movement in the soil and drainage systems will eventually affect the water table, and its seasonal variations can influence the environment and society, especially in the context of future climate changes. In Denmark, knowledge of the water table is crucial for mapping flood risk, which is of special interest during wintertime when a shallow water table can induce flooding by constraining water storage. In the agricultural context, knowledge of high and low water tables is crucial for managing the cultivation of the land. The water table in Denmark has been monitored since 1970 and data has been recorded in the national well database, JUPITER (41). The first maps of the water table in Denmark were made using the National water resources model of Denmark (DK-model). The DK-model is a dynamic hydrological model that integrates groundwater and surface water processes. The model was originally set up at a spatial resolution of 500 m, which has been updated to 100-m resolution (80–82). Recently, Koch et al. (69) mapped the water table at 10-m resolution by applying a knowledge-guided machine learning framework that was built upon simulation results from the DK-model, the Danish well database JUPITER with 13,000 wells and 19,000 groundwater proxy observations for lakes, streams and the coastline. Koch et al. (69) mapped the water table for typical winter- ( Figure 4C ) and summertime conditions. Future work should include the development of hybrid models that enhance physically based models, like the DK-model, with machine learning to model water table dynamics at high spatial resolutions while also resolving temporal dynamics more accurately (83).

Mapping the spatial variables related to water movement through soil at high resolution will aid land management. This information could reveal the potential synergies and tradeoffs between provisioning (e.g., food production) and regulating ecosystem services (e.g., climate regulation) and orient future strategies for sustainable landscapes. For instance, the drainage systems increase food production and influence the Zero Hunger goal (SDG 2), but they can also speed up the movement of nutrients to water bodies and compromise Life Below Water (SDG 14). The future challenge is to find the balance between soil functions.

Storing, filtering and transforming nutrients are some of the main soil functions and are linked to the provision of Clear Water (SDG 6). Understanding the spatial dynamics of nutrient cycling in the soil can be an important tool to improve land management, especially in Denmark where agriculture and livestock production add a large volume of nutrients to soils through synthetic fertilizers and manure. In addition, the use of pesticides in agriculture and the risk of leaching pose a challenge to groundwater quality.

In the 1980s, environmental problems, such as the eutrophication of coastal waters prompted the Danish government to accelerate research into the losses of nitrogen and phosphorus from agricultural land and to regulate the application of chemical fertilizers and animal manures on soils. The government devised national Actions Plans for mitigating N losses to the aquatic environment and implemented the National Monitoring and Assessment Programme for the Aquatic and Terrestrial Environment (62). In general, Denmark started to monitor water quality in the 1980s, and the N concentrations were measured at monitoring stations (84). The Danish Environmental Agency mapped N leaching from the root zone ( Figure 5A ). However, de-nitrification can occur and there are available maps at the national level that show this potential ( Figure 5B ). Nevertheless, the N-reduction maps are still at catchment level, and higher resolution maps would be better suited to take local characteristics that affect N-reduction into account (41). Future studies exploring the transport and reduction of nitrate in Danish landscapes at various scales, could contribute to the detailed spatial target regulation of N use. The existing DSM approaches and the availability of data at the field level could be used to create N-reduction maps at high resolution.

Unlike N, which can be leached in the soil matrix/macropores through water movement, the dynamic of phosphorus (P) is more complex and is intrinsically related to the soil mineral properties and erosion processes. One of the first attempts to identify the risk of soil P in Denmark was the development of the Danish P index on a local (85) and national scale (86). However, the Danish P index was based on a qualitative framework and the absence of absolute P losses made the application of local regulations impractical. P can be lost from the soil system by water erosion as particulate P, bound P in the macropore flow transport, and as leaching of dissolved P. Onnen et al. (87) estimated the soil loss at 10-m resolution using the soil texture maps [ Figures 3A–D (52)] and the erosion model WaTEM. Considering the average P concentration in the soil, P losses by erosion are estimated to be 56 t per ha per year (56). The potential of soil P loss through macropore transport was estimated using the map of soil P mobilization potential from 475 sites in Denmark (56) combined with the risk map of macropore transport [ Figure 5C (75)]. The P sorption capacity of Danish soils was assessed [ Figure 5D (57)] as well as the annual P leaching through the soil matrix (56), thus associating the knowledge of phosphorus binding kinetics in soil with annual drainage runoff (73) and drainage water transport in macropores and the soil matrix.

Pesticide transport is another potential environmental problem in Denmark, which various efforts have aimed to identify (88, 160). The pesticide maps mostly use data from monitoring wells, groundwater and gauging stations. A new Danish pesticide load risk indicator was developed in 2019 (159) and showed higher values for several pesticides in areas with loamy soils. This could be linked to the intense crop cultivation on the loamy Danish soils. In addition, Rosenbom et al. (90) identified a higher transport of pesticides in loamy soils compared with sandy soils, and this was associated with the transport through the macropore system, which is higher in loamy clay soils. Therefore, the association of soil macropore transport with pesticide application at a specific scale represents an important subject for future studies.

The studies on N, P and pesticide dynamics support the importance of the soil as a filter, which brings many benefits such as clean water for human needs and protection of the aquatic environments (Life Below Water - SDG 13). Historic mapping approaches proved to be efficient in orienting policy regulations, but future detailed mapping efforts can inform targeted actions by taking into consideration the local characteristics. The farmers could benefit from these targeted regulations, especially related to the use of N on croplands. In this case, assessing N dynamics at the field scale could give a better indication of the optimal N application for the local conditions. Future studies might go in this direction.

Acid sulfate (AS) soils mostly occur in wetlands located in Jutland (c. 6,500 km²). The main characteristic of wetland soils is the presence of saturated hydromorphic conditions, and farming in such areas therefore typically requires artificial drainage, which can lead to the formation of AS soils. Aeration oxidizes iron sulfides (mainly pyrite), which produces sulfuric acid that causes metals to leach and the soil pH to drop below 3. The resulting toxic combination of acidity and metals (e.g., iron and aluminium) can cause severe ecological damages to the recipient watercourses (i.e., killing fish and other aquatic organisms) and degrade concrete and steel structures located underground to the point of failure. Since small hotspots of AS soils may impact large water bodies, accurate maps are crucial for effective mitigation. The Danish legislation prohibits drainage of areas registered as potential AS soils without prior permission from environmental authorities.

A first survey of Danish AS soils was carried out between 1981 and 1984 (19). The survey comprised soil samples from approximately 8,000 locations, with pH measurements at the time of sampling and after incubation, and an estimation of pyrite content and acid-neutralizing capacity (27). All the data was then collated within the Ochre Classification database, ochre referring to characteristic orange/brown-coloured iron oxides leaching into the drainage pipes and recipient watercourses. Madsen et al. (18) presented the resulting conventional map of potential AS soils. The map classified wetland areas depending on the frequency of potential AS soils ( Figure 6A ).

More recently, Beucher et al. (91, 92) used the DSM approach for mapping the occurrence of potential AS soils. The study used the Ochre classification database and an array of environmental variables as input data. The environmental variables included a digital elevation model and derived terrain attributes, as well as legacy data on soil, landscape, land use and climate. While the first study employed Artificial Neural Networks (ANN) for prediction, the second evaluated Convolutional Neural Networks (CNN). While the ANN model already provided promising results in terms of accuracy, the CNN model was more accurate and yielded a more realistic pattern for the occurrence of potential AS soils. Figure 6 focuses on two areas of interest, the Store Vildmose raised bog area and the Skjern River catchment area located in northern and western Jutland, respectively. Particularly in the Store Vildmose area, the map generated by the ANN model showed many small sub-areas, often without natural transitions in the probability values ( Figure 6B ). In contrast, the predictive map created with the CNN model shows more logical transitions and patterns ( Figure 6C ). The CNN-based map also corresponds more closely to the conventional map ( Figure 6A ) than the ANN-based map, which can also be noted in the Skjern River catchment area. This consistency can be explained by the fact that CNN enables the inclusion of spatial contextual information extracted from environmental covariates around each soil observation used in the model as input data. Therefore, the resulting predictive map renders a more accurate representation of the occurrence of potential AS soils than the ANN-based model. Despite its lack of detail, the conventional map was developed using solid expert knowledge, which corroborates the CNN-based map. The predictions of the best-performing CNN model could also be elucidated using the model-agnostic interpretation method SHapley Additive exPlanations (SHAP) which outlined the most important covariates in terms of contribution to the model, not only quantitatively, but also geographically [ Figure 6D (92)].

One of the main challenges of the 21^st century is to increase food production to meet the projected population growth (93). However, the expansion of agriculture over natural areas is not a sustainable solution, and the focus should therefore be to manage the land in a way that optimizes the usage and conservation of resources. Denmark has a long history of collecting information about land management and crop yields, and since 2011 the land use information has been available at the field level (94). The specific land use and management are typically the results of multiple factors, varying from environmental, economic, social and political factors. Understanding the factors that drive land use change, agricultural land suitability and crop yields is an important step to improve the use of natural resources (e.g., soil). Advances in remote sensing, mapping and modelling approaches have generated new possibilities for combining these factors and providing detailed spatial information for stakeholders.

Denmark has historical data about crops from the years 1644 and 1844 at parish level and high-resolution land use maps from 1990 to the present (38). The methodology used to create the recent land use maps is based mainly on the combination of a topographical database, field parcel maps from 2014 to 2020 and information about wetlands. In general, the land use change between 1990 and 2020 was characterized by a relative decrease of 6.5% (195,000 ha) in cropland areas, and increases in grasslands (26,000 ha; 18.6%), forests (96,800 ha; 17.8%), wetlands (partly water-covered; 19,000 ha; 37.9%) and surface water bodies (6,000 ha; 11.6%) (38). The study presented the maps in a raster format at a 25-m resolution. Furthermore, every year Danish farmers register the field crops used in order to receive support from the Common Agricultural Policy of the EU, which has provided near-complete field level information on cropping patterns in Denmark since 2011 (https://eng.lbst.dk/). However, even more detailed information within the land use polygons could be made available. Remote sensing images and advances in machine learning could potentially help to produce maps that present not only the vegetation class but also the status of the vegetation. For instance, supervised land use classification could potentially map different levels of degradation in grasslands. This additional information could be useful for farmers to monitor the crops, but also for policy-makers for targeted regulations and subsidies.

The observed land use generally depends on soil characteristics, climate, relief, the water table, as well as economic, social and political factors. Placing the most suitable crop in the right environment would not only favour productivity; in many cases it would also be the most sustainable option. Laville (95) coined the term “Natural Terroir Unit” (NTU) to define locations that share the same topography, soil and climate, providing a unique high-quality characteristic for an agricultural product. Several studies have applied DSM methods to identify NTUs using environmental characteristics in Denmark (96–98) and around the world (99, 100). Importantly, Peng (98) noted a missing link between the existing terron maps and agricultural data. The authors addressed this issue by comparing their terron maps ( Figure 7B ) with historical wheat yield ( Figure 7C ). This allowed them to assess the validity of the terrons and their relative productivity for a specific crop.

It is also possible to predict the agricultural land suitability with crop growth models, such as EcoCrop (101), and species distribution models (94, 102) mapped the land suitability for specialty crops in Denmark ( Figure 7D ) using EcoCrop and machine learning models based on MaxEnt. They found that the advantage of machine learning models was their ability to include explanatory variables related to economic and social factors, which are intrinsic to the land use. Selecting the appropriate crop for a certain location would minimize risks from adverse weather, pests, diseases, and changes in market conditions and policies. However, the authors also stressed that maps need to have a clear and unambiguous interpretation in order to be useful for end users.

The need for selecting healthy sites is crucial when planting trees, especially when wood production is the goal. Denmark intends to double the 1989 baseline forest cover within 80-100 years, and due to the longevity of forest stands, not only does the present site quality need to be considered to ensure healthy trees as climate shifts, but so does also the future site quality. One way to evaluate forest site quality for various species is to use ecograms. Ecograms are used to determine site quality for various tree species by determining the nutrient and water availability at a site, and comparing them to the requirements for the specific species. Figure 7A shows the spatial distribution of nutrient and water supply in Danish soils (103). To build this map, the landscape was classified into six nutrient classes and nine water classes based on four variables: pH at 1 to 2 m depth, average precipitation between April and October, groundwater depth, and plant-available water. Since the 1990s, Denmark has used ecograms for small-scale implementation when assessing sites for various tree species (29), and for the implementation of close-to-nature forest management. Recently, national ecogram maps were generated for five species to help foresters quickly determine which regions are suitable or unsuitable for the different species for future afforestation efforts (104).

Machine learning is also a potential means to map crop yields. The availability of historical and current wheat yields in Denmark made it possible to model the potential and actual yield through time (105). The authors found that advances in agricultural management, including fertilization, pest management, irrigation and drainage allowed farmers to exceed the historical yield predictions. Larsen et al. (106) also modelled the potential yield of the energy crops willow (Salix) and silver grass (Miscanthus) across Denmark. Their results showed that willow yields were likely to be higher than silver grass yields on sandy soils, whereas silver grass yields were highest on clay-rich soils. This information could be important for selecting feedstocks for bioenergy.

Agricultural production is one of the most generally recognized soil functions, and studies related to land suitability and crop yields highlight the undeniable importance of the soil. The growing conditions for most crops depend very strongly on the soil. Therefore, outlining the optimal environments for specific crops has the potential to improve the use of local resources and reduce the severity of the risks that farmers face. It also helps to provide a clear link between food production and soil functions. These efforts therefore help to improve and maintain agricultural production in Denmark. This is highly relevant in Denmark, which produces three times more food than the local population consumes (79), which contributes to the progress toward Zero Hunger (SDG 2).

Peatlands are unique, organic, wetland ecosystems and very important for climate change mitigation and the environment. Global peatland coverage is estimated to be about 3% of the total land area only, but they contain about 5-20% of the global soil C stock (107, 108). Peatlands provide several ecosystem services that contribute to several SDGs, such as Climate Action - SDG13 (C sequestration), Life on Land - SGD15 (biodiversity conservation), Clean Water - SDG 6 (hydrological regulation). However, human-mediated degradation due to poor management impairs the delivery of many peatland ecosystem services (109–111). Under refined management systems, however, peatlands may offer opportunities to achieve key global sustainability objectives (112).

In Denmark, peatlands cover approx. 2,400 km² ( Figure 8 ). Various peatland inventories have been established through time. The first real peatland mapping attempt on a national scale commenced in 1919 (focusing specifically on bogs). This was to ascertain the quality and potential use of bog resources for fuel due to failed supply of foreign fuel during World War I and II (113). From the early 1920s, the Danish Land Development Service classified meadows and bogs larger than five hectares into four groups depending on the quality of the peat for fuel (113, 114). Between 1975 and 2011, several national soil databases including peatland soils were established: the DSC, the Ochre Classification, and the SINKS databases. The original SINKS database resulted from a national field survey carried out in 2010 and 2011. About 9,800 soil samples were collected following a systematic sampling scheme (using three regular grids of 250-, 275- and 500-m spacing: Figure 8A ). At each sampling location, four subsamples were taken at 30-cm depth increments from the soil surface down to 120 cm depth. Furthermore, other detailed inventories have been established at field scale for peatlands of interest.

To realize national commitments to international climate agreements (i.e., Kyoto Protocol, Paris Agreement), accurate mapping of peatlands is required. Thus, the detailed mapping of peat properties is needed; conversely, this is also difficult to achieve using solely conventional surveys due to the challenges of time, labour input, and inaccuracies in measurements (115, 116). To this end, DSM techniques have been employed in Danish peatland studies. Generally, DSM techniques enable the mapping of peatland properties and processes such as the extent, diminishing C stocks, response to climate change and the impacts of rapid land use changes more accurately (107). At the national scale, both the distribution of SOC and changes in peatland extent were mapped by Greve (31) and Kheir et al. (117) using geostatistical and machine learning techniques, respectively. Moreover, at field and farm scales, proximal geophysical sensors, which are known to provide opportunities to rapidly acquire more accurate ancillary data due to the unique electrical properties of peat (e.g., water content, electrical conductivity, relative dielectric permittivity), have been used (115, 118). Specific applications of these techniques include studies by Knadel (119) and Beucher (120), who employed not only spectroscopic techniques for SOC mapping but also electromagnetic induction techniques for peat thickness mapping, both at field scale. Recently, a policy support report by Møller et al. (121) described potential methods for an improved mapping of SOC in peatland areas. The study also outlined the possible improvements in accuracy that would be gained from the use of geophysical soil sensors.

The various soil mapping activities have provided valuable insights on the functioning and status of Danish peatlands. For example, Beucher et al. (120) could accurately predict peat thickness at field scale (from apparent electrical conductivity and other relevant predictors), which is vital for estimating C stocks and meeting binding climate policies such as the EU 2030 Climate and Energy Framework. Additionally, the effect of drainage and historical peat extraction for fuel by humans was quantified by Greve et al. (31) as a 35% reduction in the national extent of peatland areas between 1975 to 2010. To halt the further decline in the peat coverage, this finding was significant for stakeholder action. Similarly, the human-induced subsidence of the Store Vildmose raised bog ( Figure 8B ) over a 130-year period revealed near complete loss of peat in extracted areas and subsidence of at least 2 m along the greater part of an established transect. In contrast, areas with peat growth (2-6 mm per year) were also observed and this was due to laws that protected such areas and thereby prevented the drainage and degradation of the peat (122). Currently, several peat mapping activities are underway to restore the Store Vildmose peatland to its natural state. This includes plans to stop agricultural activities and to rewet the area. The purpose of the lowlying project is to reduce CO₂ emissions to promote the natural qualities of the area, to improve coherence and robustness of natural areas and to restore more natural hydrology (123).